In the case systems of English and other Indo-European languages, there's typically a subject (agent of the action), a verb (action) and an object or several objects (patient(s) of the action). Trigger languages on the other hand divide things into agent, patient and other objects.
The differences between "trigger languages" and "nominative/accusative languages" are:
- What the sentence is about (the emphasis so to speak) is lexicalized, although still pragmatic -- This is called the trigger, focus or topic of a sentence.
- The trigger can be the agent or the patient or perhaps have any other theta role, so calling the triggered argument "subject" does not make sense.
- Such languages need no explicit passive voice because they can do this with triggers: for instance,
- I-TRG drink-TRG=AGT milk-PAT (I drink milk.)
- I-AGT drink-TRG=PAT milk-TRG (I drink milk. / Milk is drunk by me).
Triggering works as follows: Every argument of the sentence is marked for its role (agent, patient, other objects). The triggered argument is marked on the verb. In Carsten Becker's conlang Ayeri, perhaps the best known trigger conlang, trigger and case marker change places.